Comparative analysis of structural representations of images, based on the principle of representational minimum description length

摘要

Based on the representational-minimum-description-length (RMDL) principle, proposed earlier and intended for the quantitative estimate of the degree of invariance of image representations, a comparative analysis has been carried out of several segmentation algorithms that construct contour descriptions of images, as well as several algorithms for constructing structural descriptions of images by using various alphabets of structural elements. It is shown that it is adequate to use the RMDL principle when comparing the invariance of the representations of images. It is established that the optimum segmentation algorithms differ for different samples (for example, aerospace images or images obtained indoors). Based on an objective criterion, it is shown for the first time that it is expedient to use straight line segments and segments of second-order curves as structural elements and that higher-order curves have low efficiency. (C) 2008 Optical Society of America.